An interaction effect analysis of thermodilution-guided hemodynamic optimization, patient condition, and mortality among successful cardiopulmonary resuscitation patients

Post-cardiac arrest syndrome and target temperature management have an inuence on hemodynamics in a complex way. Proper hemodynamic monitoring is necessary among post-cardiac arrest patients. The aim of our study was to investigate the effects of PiCCO™-guided hemodynamic management on short-and long-term mortality in post-resuscitation therapy and to disentangle the relationship between PiCCO™-application, the patients' condition, and mortality by assessing it as an interaction effect. this longitudinal analysis of comatose patients after successful cardiopulmonary resuscitation to 30 received no and retrospective longitudinal chart study of after Institutional medical records and charts were reviewed to estimate the in-hospital management in the rst 72 after admission. Information related to pre-hospital emergency care by OHCA patients was obtained from the records of emergency medical service and Utstein reports.


Introduction
Physicians face several challenges during post-cardiac arrest therapy at the intensive care unit (ICU).
Post-cardiac arrest syndrome occurring after the return of spontaneous circulation (ROSC) contains reperfusion-ischemic injury, post-cardiac arrest brain injury, post-cardiac arrest myocardial dysfunction and the precipitating pathology itself that caused cardiac arrest. 1 The prevention and treatment of secondary brain damage and hemodynamic management are co-dependent and both elements play an important role in post-cardiac arrest therapy. 2 On the one hand, target temperature management (TTM) with target temperature of 32-36 ºC is applied as a neuroprotective tool in comatose patients surviving cardiac arrest. [3][4][5] On the other hand, TTM may in uence circulatory system in a negative manner leading to bradycardia, arrhytmias, increased systemic vascular resistance and hypovolaemia. 6,7 Post-cardiac arrest syndrome has also a plenty of factors affecting the hemodynamics harmfully. 1 Moreover, a proper hemodynamic management is crucial to keep a satisfactory cerebral perfusion to prevent further cerebral deterioration. 8 Therefore, advanced hemodynamic monitoring may be useful in post-resuscitation therapy. However, there is no evidence which hemodynamic parameters should be monitored and which monitoring method should be used.
The PiCCO™ (Pulse index Contour Cardiac Output) monitoring system (Pulsion Medical Systems, Munich, Germany) guided therapy uses transpulmonary thermodilution and pulse contour analysis to determine hemodynamic parameters. 9 A cold liquid bolus is applied via a central venous catheter passing through various thoracic compartments, and a peripheral arterial thermodilution catheter (mostly inserted into femoral artery) detects the temperature curve, calculating cardiac output with the modi ed Steward-Hamilton equation. 10 In addition, a continuous cardiac output monitoring by recording the pulse pressure wave can also be used in PiCCO™. Calculated hemodynamic parameters give further information about the circulatory system: the global end diastolic index (GEDI) re ects preload and systemic vascular resistance index (SVRI) represents afterload, while the extravascular lung water index (ELWI) gives information about pulmonary edema, and global ejection fraction (GEF) re ects cardiac systolic function.
PiCCO™ was shown to provide better monitoring capacities in critically ill patients with necrotising pancreatitis, 11 sepsis 12 and in cardiac surgery. 13 It was also assessed during post-cardiac arrest therapy and therapeutic hypothermia (32-34 ºC) showing that thermodilution measurements are precise in patients following ROSC, even if lower temperature is used. 14 In addition, a set of triplicate injections was recommended to be applied in order to achieve the best precision. 14 As such, the European Society of Intensive Care Medicine recommends the use of cardiac output monitoring (and transpulmonary thermodilution) in patients with severe shock not responding to initial treatment. 15 However, to our knowledge there is no assessment regarding whether PiCCO™ monitoring guided therapy actually improves mortality outcome during post-resuscitation therapy.
Therefore, the aim of our study was to investigate the effects of PiCCO™-guided hemodynamic management on 30-day and one-year mortality in comatose patients treated with TTM after ROSC. More speci cally, we focused on disentangling the relationship between PiCCO™-application, the patients' condition, and mortality by assessing it as interaction effects.

Study design
This was a retrospective longitudinal chart review study of patients (N = 254) who successfully underwent cardiopulmonary resuscitation after either OHCA (out-of-hospital cardiac arrest) or IHCA (inhospital cardiac arrest) at Semmelweis University Heart and Vascular Center between January 2008 and January 2015. Institutional medical records and charts were reviewed to estimate the in-hospital management in the rst 72 hours after admission. Information related to pre-hospital emergency care by OHCA patients was obtained from the records of emergency medical service and Utstein reports.
The Semmelweis University Regional and Institutional Committee of Science and Research Ethics approved our study (approval number: 19/2019). The written informed consent requirement was waived due to the retrospective nature of the study.

Patients
We included in our analysis those comatose patients who were cooled to 32-34 °C for 24 hours after ROSC on the basis of the even actual ERC (European Resuscitation Council) guidelines, 16 were older than 18 years, had no end stage illness in history, were not pregnant, had no active bleeding, whose cause of cardiac arrest had a probable cardiac origin, and were not involved in a clinical trial. In addition, only patients cooled with Blanketrol III™ (Cincinatti SubZero Products, Cincinatti, USA) thermo-feedback device were enrolled into the study -those patients whose temperature management was applied with ice packs and/or physical cooling were excluded, because target temperatures could not be reached in most of these cases. 23 patients were excluded from the study due to lack of data. After the sampling process depicted in Fig. 1, the nal study sample included 63 patients (33 who were applied and 30 who were not applied PiCCO™ monitoring).

Initial therapy
Post-resuscitation therapy and TTM were initiated as soon as possible after the admission of OHCApatients and after ROSC of IHCA-patients. All patients received standardized critical care according to our institutional protocol. An acute coronarography and percutanous coronary intervention (PCI) were performed and intra-aortic balloon pump (IABP) was inserted, if indicated. All patients were treated in the ICU at the acute phase of the assessment. We upheld mechanical ventilation until the patients ful lled extubation criteria.
Oxygen saturation, electrocardiogram, invasive arterial blood pressure, central venous pressure, diuresis, blood gas parameters and serum lactate level were monitored for all patients. An echocardiography was performed after the admission to the ICU to assess the heart function. If it was available, the hemodynamic monitoring was augmented with PiCCO™ monitor (Pulsion Medical System, Munich, Germany); thermodilution measurements were applied at least every 12 hours during the rst 48 hours after the initiation of cooling. We measured the following variables: cardiac index (CI: l/min/m 2 ), systemic vascular resistance index (SVRI: dyn.sec.cm − 5 ), global end-diastolic volume index (GEDI: l/m 2 ), extravascular lung water index (ELWI: ml/kg/m 2 ) and global ejection fraction (GEF: %). If patients had IABP their device was paused for the time of thermodilution measurement.
Fluid, vasopressor, and inotrope therapy were accomplished by monitoring heart rate (HR), mean arterial pressure (MAP), central venous pressure, diuresis and lactate levels for patients without PiCCO™. We used the goal parameters given by the actual ERC guidelines. 16 The hemodynamic management was guided by PiCCO™ parameters and the decision tree of Pulsion Medical System 17 was applied for PiCCO™ patients.
Noradrenaline, dobutamine, dopamine and/or adrenaline were used as vasopressors and positive inotropic agents in mono-or combination therapy. If the cardiac function was critically impaired, a 24hour long levosimendan treatment was implemented. The measured hemodynamic parameters determined which drug was chosen during the assessment based on the above mentioned therapy decision tree of Pulsion Medical System. 17 TTM was divided into three phases: induction of cooling, maintenance phase, and rewarming. During the induction of TTM 30 ml/kg crystalloids were given to the patients, and the previously mentioned thermofeedback device was used. The temperature in maintenance phase was upheld with the same device. After 24 hours of hypothermia, rewarming was achieved by passively trying to keep a 0.25˚C/hour rewarming speed until normothermia was reached. Prevention and control of fever still played an important role in the rst 72 hours of the assessment. Patients received a combination of intravenous benzodiazepine and opioid as sedation during hypothermia. Their core temperature was measured with an esophageal thermometer.
Additionally, patient data extracted included the patients' age, gender, previous health conditions affecting cardiovascular system (diabetes, hypertension, stroke, hyperlipidemia, and acute myocardial infarction), circumstances of cardiopulmonary resuscitation (time between collapse and ROSC, if the patient was on monitor at the time of collapse, if basic life support was performed, and initial rhythm) and important steps of therapy after ROSC (presence of cardiogenic shock (CS), ST segment elevation and non-ST segment elevation myocardial infarction (STEMI and NSTEMI), frequency of acute PCI, frequency of IABP insertion, ejection fraction after ROSC, and necessity of levosimendan administration after ROSC).
Primary outcome was de ned as mortality after 30 days and secondary outcome was de ned as mortality after one year.
Patients` changes in temperature during TTM, hemodynamic parameters, changes in serum lactate levels during TTM, and catecholamine dosing can be found in Supplement 1.

Statistical analysis
First, three sets of bivariate statistical tests were performed: we assessed differences between 1. the PiCCO and non-PiCCO groups (yes vs. no), 2. mortality after 30 days (yes vs. no) and 3. mortality after one year (yes vs. no). Mann-Whitney's U tests was used for continuous variables and Chi-square test or (in case of small sample sizes) Fisher's exact test was applied for categorical variables. Additionally, Kaplan-Meyer curves and Log-rank tests for signi cance were used as longitudinal data to asses differences in mortality between PiCCO and non-PiCCO groups and for those variables where mortality was at least marginally signi cant using the categorical Chi-square analysis. Second, interaction effects were explored (candidate variable vs. PiCCO™ use vs. mortality) using the same statistical methods as in the bivariate analysis for those variables that had at least marginal associations with both PiCCO™ use and mortality. Additionally, logistic regression analysis was performed using the interaction terms as dummy variables. If there were zero cell sizes, then dummy categories were combined with non-zero cells. In a rst set of logistic regression models, all interaction dummy variables were included, and in a second set of models only statistically signi cant dummies stayed.
Continuous variables are described with median values and their corresponding interquartile range, and categorical data are described as percentages. Not more than 10% of the data were missing; we performed multiple imputation using the k-nearest neighbor algorithm to replace variables with missing values.
We identi ed p < 0.05 for statistical signi cance and p < 0.2 and p ≥ 0.05 for marginal signi cance. Data management and statistical analysis was performed using TIBCO STATISTICA v13.4 (Tibco Software Inc., Palo Alto, CA), and gures were created using GraphPad Prism version 5.0 (GraphPad Software, La Jolla, CA).

Results
Characteristics of patients, PiCCO™ use and 30-day and 1year mortality Patient characteristics are shown in Table 1. Altogether, 52% received PiCCO™, 38% died after 30 days, and 57% died after one year. Patients with PiCCO™ application were signi cantly more likely to die after 30 days, and marginally more likely to die after one year than non-PiCCO™ patients (Fig. 2). Interaction effects between 30-day mortality, PiCCO™ use and patient characteristics Figure 4 visualizes the multivariate interaction effects between 30-day mortality, PiCCO™ use, and patient characteristics, statistically controlled for the effects of the subgroups. Accordingly, patients with either PiCCO™ or hyperlipidemia were marginally and patients with both PiCCO™ and hyperlipidemia were signi cantly more likely to die at day 30 than patients with neither PiCCO™ nor hyperlipidemia. Moreover, patients with PiCCO™ but no STEMI were signi cantly more likely to die than patients with no PiCCO™ or patients with both PiCCO™ and STEMI. Additionally, patients with cardiogenic shock regardless of PiCCO™ were signi cantly much more likely and patients with PiCCO™ but not cardiogenic shock were signi cantly more likely to die than patients with neither cardiogenic shock nor PiCCO™. Furthermore, patients with PiCCO™ but no IABP were marginally and patients with IABP regardless of PiCCO™ were signi cantly more likely to die than patients with neither IABP nor PiCCO™. In addition, patients receiving catecholamine treatment after ROSC but no PiCCO™ were marginally more likely and patients with both catecholamine and PiCCO™ were signi cantly more likely to die than patients with no catecholamine. Moreover, patients with no stroke but with PiCCO™ application were signi cantly more likely to die than patients with neither stroke nor PiCCO™ or patients with both stroke and PiCCO™. Furthermore, higher mortality was seen in patients with PCI but without PiCCO™ compared to patients without PCI regardless of PiCCO™ or patients with both PCI and PiCCO™.

Interaction effects between one-year mortality, PiCCO™ use and patient characteristics
Multivariate interaction effects between one-year mortality, PiCCO™ use and patient characteristics, statistically controlled for the effects of the subgoups, are shown in Fig. 5. Patients with PiCCO™ but no STEMI were marginally more likely to die than patients without PiCCO™ regardless of STEMI or patients with both PiCCO™ and STEMI. Additionally, patients recieving catecholamines after ROSC regardless of PiCCO™ were signi cantly more likely to die than patients who did not receive catecholamine treatment.
Moreover, patients with PiCCO™ application but without stroke in past history were marginally more likely to die than patients with both stroke and PiCCO™ or patients with neither stroke and nor PiCCO™. Additionally, although no interaction was found for CS, PiCCO™ and one-year mortality, and IABP, PiCCO™ and one-year mortality, CS and IABP were independent predictors of mortality. Finally, once controlled for subgroups, the interactions between hyperlipidemia, PiCCO™, and one-year mortality, and between PCI, PiCCO™, and one-year mortality were no longer statistically signi cant.

Discussion
To our knowledge, there are no published studies regarding the association between PiCCO™ guided therapy and survival in post-cardiac arrest treatment. We found ve interaction patterns between patients' condition and being monitored with PiCCO™ with regard to mortality after 30 days, which distilled down to three interaction patterns by the end of the rst year. As such, our analysis presents a valuable insight into the nuances of advanced hemodynamic monitoring and hemodynamic management during postcardiac arrest therapy.
The uniqueness of our study is that we disentangled the interaction effects between PiCCO™ application, mortality and patients' condition in order to elucidate the potential cause of decayed survival rate in PiCCO™ monitored patients. We found that there was a complex interaction between the use of PiCCO™ and both 30-day and 1-year mortality, depending on the condition of the patient.
We identi ed ve groups regarding 30-day mortality. In the rst group having a condition and receiving PiCCO™ meant increased risk compared to either not having the condition or not receiving PiCCO™, as in the case of hyperlipidemia. Speci cally, 30-day mortality in patients who both had hyperlipidemia and also received PiCCO™ was higher than among those who either had no hyperlipidemia or received no PiCCO™.
In the second group not having the condition but receiving PiCCO™ meant increased risk compared to either not having the condition or having the condition regardless of PiCCO™, as in the case of STEMI and stroke. Speci cally, 30-day mortality in patients without STEMI (without stroke, respectively) who also received PiCCO™ was higher than among those who either had no STEMI (no stroke, respectively) and received no PiCCO™, or had STEMI (stroke, respectively) regardless of PiCCO™.
In the third group having the condition or receiving PiCCO™ meant increased risk compared to not having the condition and not receiving PiCCO™, as in the case of CS and IABP. Speci cally, 30-day mortality in patients who had either CS (IABP, respectively) or received PiCCO™ was higher than among those who neither had CS (IABP, respectively) nor received PiCCO™.
In the fourth group not having the condition or receiving PiCCO™ meant increased risk compared to having the condition and not receiving PiCCO™, as in the case of PCI. Speci cally, 30-day mortality in patients who had either no PCI or received PiCCO™ was higher than among those who had PCI and did not receive PiCCO™.
In the fth group having the condition regardless of PiCCO™ meant increased risk compared to not having the condition, as in the case of recieving catecholamine. Speci cally, 30-day mortality in patients who received catecholamine was higher than among those who did not receive catecholamine, regardless of PiCCO™.
Furthermore, we identi ed three groups regarding one-year mortality. In the rst group, there was no difference in one-year mortality either regarding PiCCO™ application or regarding the presence of the condition, as in the case of hyperlipidemia in prior history or treatment with PCI after ROSC.
In the second group, having the condition regardless of PiCCO™ meant increased risk compared to not having the condition, as in the case of IABP, CS and catecholamine administration. Speci cally, the oneyear mortality was higher in patients treated with IABP (having CS, recieving catecholamine, respectively) than among patients without these conditions regardless of PiCCO™.
In the third group, not having the condition but receiving PiCCO™ meant increased risk compared to either not having the condition or having the condition regardless of PiCCO™, as in the case of STEMI and stroke. Speci cally, one-year mortality in patients without STEMI (without stroke, respectively) who also received PiCCO™ was higher than among those who either had no STEMI (no stroke, respectively) and received no PiCCO™, or had STEMI (stroke, respectively) regardless of PiCCO™.
Adequate hemodynamic management is one of the key elements in post-cardiac arrest therapy. Patients after successful cardiopulmonary resuscitation may experience a global ischemic-reperfusion injury and myocardial depression, as parts of post-cardiac arrest syndrome, leading to hemodynamic instability after ROSC. 1,18 Furthermore, the precipitating cause of cardiac arrest itself may result in a deterioration of hemodynamic parameters. It is also well-known that TTM has an in uence on hemodynamics by several pathways. A decrease in heart rate and cardiac output, an increase in systemic vascular resistance, and hypovolaemia caused by raised diuresis may be present as consequences of lower body temperature. [19][20][21] As the path leading to hemodynamic instability in post-cardiac arrest syndrome is multifactorial, there is a requirement of proper monitoring tools and proper hemodynamic goal parameters to guide the therapy of these patients.
Despite the complexity of the circulatory effects in post-cardiac arrest period, there is no clear evidence and therefore guidelines about exactly which parameters should be monitored, which goal parameters should be kept during patients' management, and which monitoring tools should be used to guide the treatment. [22][23][24][25][26] The current ERC guidelines recommend to target a mean arterial pressure to achieve a satisfactory urine output (1 ml/kg/h) and a decreasing or normal serum lactate level, considering the patient's habitual blood pressure, the cause of cardiac arrest, and the severity of probable cardiac dysfunction. 2 Moreover, the guidelines suggest that additional cardiac output monitoring may help to guide therapy in hemodynamically unstable patients. 2 However, there is no evidence that cardiac output measuring affects outcome in this patient group.
Transpulmonary thermodilution and pulse contour analysis are applied during PiCCO™ monitoring, which allow intermittent and continuous measurements of cardiac output. Furthermore, several additional measured and calculated parameters can be estimated beside cardiac output re ecting preload, cardiac function and the vascular tone. 9,27 However, artery pulmonary catheter is the gold standard of cardiac output measurement. 28 PiCCO™ is less invasive, it is less in uenced by respiratory uctuations, and has a longer dwell time. 29,30 The European Society of Intensive Care Medicine suggests the application of artery pulmonary catheter only in refractory shock with concomitant right ventricular failure. 15 PiCCO™ was found to be effective in the evaluation of hemodynamic situations in critically ill patients, leading to a faster decision making. 31,32 The PiCCO™-guided hemodynamic management shortened the duration of vasoactive therapy, mechanical ventilation and ICU stay among elderly patients with cardiogenic shock after acute myocardial infarction. 33 However, there is a lack of evidence regarding the effect of PiCCO™ monitoring system and PiCCO™ guided therapy on mortality in post-cardiac arrest treatment -and this is why our study ndings are unique and much needed.
The results of the interaction effect analysis between PiCCO™ application, mortality, and patients' characteristics show that more severe patient condition per se was not the cause of higher mortality rate in the PiCCO™ group. Moreover, patients in better health conditions (without STEMI, without cardiogenic shock, without IABP device or without stroke in prior history) had worse outcomes in PiCCO™-guided therapy. Our nding supports the literature, as the European Society of Intensive Care Medicine recommends the use of cardiac output monitoring and the application of transpulmonary thermodilution in patients with severe shock not responding to initial therapy. 15 Furthermore, we need to point out that the injured brain after successful CPR frequently has an impaired autoregulation resulting the MAP dependence of cerebral blood ow. 34,35 This fact raises the question if the monitoring of cardiac output and advanced hemodynamic parameters is the proper way to guide post-cardiac arrest therapy or the simpler MAP measurement and the observation of tissue perfusion give enough information.
We suggest applying PiCCO™ in post-cardiac arrest therapy in selected cases based on our results.
Moreover, further prospective studies are needed to clarify which patient groups bene t from cardiac output monitoring and thermodilution methods after successful CPR.
Another important nding of our study was that catecholamine administration worsened both 30- Our study has a number of limitations, most of which are related to its retrospective nature and small sample size. Patients were enrolled into the study non-randomly, and the analysis was based on retrospectively collected data. The grouping of patients was assigned based on the availability of a PiCCO™ monitor. However, baseline characteristics were well-balanced although due to performing only an interaction effect analysis between three variable groups but not further being able to adjust for other factors may have left some element of confounding.
The improper evaluation of PiCCO™ measurements should represent an additional limitation of our study.
However, the trends of catecholamine and vasoactive dosage followed the changes of speci c hemodynamic parameters, showing the adequacy of therapy. In addition, serum lactate levels decreased in parallel during and after TTM in both the PiCCO™ and non-PiCCO™ groups.

Conclusion
Post-cardiac arrest syndrome may lead to hemodynamic instability caused by multiplex factors, including the effects of TTM. Proper hemodynamic monitoring and management of hemodynamic parameters is indispensable in this patient group. The accuracy of the PiCCO™ monitoring system was con rmed in the lower temperature environment of post-cardiac arrest therapy. However, little is known about the effectiveness of PiCCO™ regarding mortality outcomes in post-cardiac arrest patients. Our analysis showed that while there was an interaction effect between PiCCO™-guided therapy, patients condition and mortality, after 30 days for most conditions and after one year we saw either no effect of PiCCO™ Availability of data and materials All data generated and analysed during this study are available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests.

Funding
No external funding was applied to this study.
Authors' contributions EK contributed in study design, acquisition, analysis, and interpretation of data and she wrote the paper.
VAGy was involved in study design, data analysis, interpretation and statistics, and revised the draft. DP, AFGy and ZsSzT contributed to study design, acquisition and interpretation of data, and they were involved in drafting the manuscript.
LG, BH, JG and BM were involved in study design, analysis and interpretation of data, added new insights to the topic and drafted the manuscript. EZ contributed to study design, analysis and interpretation of data, helped to discuss all the results and drafted the manuscript.
All authors read and approved the nal manuscript and agreed to be personally accountable for the work.  Figure 1 The selection of study population and eligibility of patients. 254 patients after successful